Improve KubeAI autoscaling support

[eero-t]

Description

KubeAI updates:

• Autoscale 8b Gaudi model
• Change 70b Gaudi model to scale-from-zero, to avoid idle pods reserving 4 devices
• Add example nodeSelectors
• Update README / fix typos in it

Issues

n/a.

Type of change

• New feature (non-breaking change which adds new functionality)

Dependencies

n/a.

Tests

Manual testing for scaling / changes.

[eero-t]

@poussa minReplicas/maxReplicas are chosen so that one could run these and some additional model(s) on a single 8x Gaudi node, if not all of them are in use at the same time. IMHO this is good default for examples like this.

[Copilot]

Pull Request Overview

This PR improves KubeAI autoscaling support by updating autoscaling configurations and providing clearer examples for node selection and model deployment. Key changes include:

• Adding commented-out example nodeSelectors to the OPEA values file.
• Updating the 70b Gaudi model configuration to scale from zero with explanatory comments.
• Adjusting the 8b Gaudi model configuration and updating the README with corrected deployment instructions.

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
kubeai/opea-values.yaml	Added example nodeSelector blocks
kubeai/models/llama-3.3-70b-instruct-gaudi.yaml	Changed autoscaling parameters and added clarifying comments
kubeai/models/llama-3.1-8b-instruct-gaudi.yaml	Updated environment variable formatting and set maxReplicas
kubeai/README.md	Revised documentation to reflect updated autoscaling behavior

[poussa]

LGTM

[eero-t]

@mkbhanda OK to merge?

[eero-t]

@marquiz, @mkbhanda OK to merge?

[marquiz]

Thanks @eero-t, looks good to me